About Acatian
Acatian develops process engineering software for chemistry, biotechnology, food and related industries. Our goal is to shorten the path from process data to a better scale-up or plant decision.
Engineering method
A process model should make its inputs, equations, units and operating limits available for review. Saved versions keep the basis of a calculation available when assumptions change.
Our public reference case includes downloadable inputs, results and numerical checks. It demonstrates reproducible software execution. Predictive accuracy must be established separately for the process and intended use.
Development direction
Less time rebuilding models. More useful experiments. Better use of production equipment, energy and raw materials.
Our direction is to train approximations on relevant process calculations and measurements, then test accuracy on cases excluded from training. An approximation must stay within its demonstrated operating domain.
We aim to connect molecular and formulation descriptors with measured endpoints such as conversion or productivity. Molecular similarity alone does not establish yield, substitutability or safe operating conditions.
We aim to compare calibrated model predictions with plant measurements and evaluate bounded control recommendations. Sensors, control systems and equipment actions need a separately implemented and qualified integration.
Evidence
Passing a software test is one part of evaluating an engineering result.
Versioned inputs, reproducible runs, numerical checks and file integrity. Demonstrated by the synthetic reference case.
Calibration and independent validation data over the intended operating range. Established for each application.
A qualified external solver or implemented site integration, with convergence, security and comparison evidence. Scoped separately from browser screening.
Named acceptance criteria, uncertainty, limitations and qualified approval for the selected model version.
Technical evaluation
Begin with one company-owned process. Agree which outputs to compare, set tolerances and review deviations before extending the study.